[12:41:39] are possible BEANS issues fairly safe to bring up here? [12:46:49] yes [12:47:20] but if they're really poisonous beans I think there's security@wikimedia.org [12:49:37] but consider that the channel is logged... [12:53:05] ;) [12:53:31] does SUL autocreate accounts on local projects if the IP address is blocked for account creation for the IP? [12:53:59] on that specific local project that is [13:01:13] no [13:02:58] or at least, I think it doesn't because I heard of people not managing to autocreate accounts, but there's https://bugzilla.wikimedia.org/show_bug.cgi?id=11148 [13:03:45] and there's https://bugzilla.wikimedia.org/show_bug.cgi?id=17929 , more importantly [13:15:20] thanks [13:16:01] FooBarMartijn, if you could do some testing it would be appreciated ;) [13:16:05] (for the former) [13:16:20] Nemo_bis, I'll take a look what I can do and test [13:16:25] good [13:16:39] being lazy, has anyone compiled a list of testcases (yet?) [13:17:16] fortunately, I have an admin flag on en. so I can experiment with my own IP [13:19:35] only very slightly related, I'm looking for where the display of block messages is called to fix a minor personal annoyance, and also get a little familiarity with the code: https://bugzilla.wikimedia.org/show_bug.cgi?id=35090 [13:19:48] where should I roughly be looking? [13:21:13] If you want the message identifier append ?uselang=qqx (hm, but will it work) [13:21:30] * Nemo_bis knows nothing about the code [13:23:38] cheers. I'll just go try and reproduce 11148 then, see if it can be closed [13:52:12] Nemo_bis, I closed the bug as fixed, per the testcase [13:52:30] yay :) [13:54:29] is it worth filing a bug for the tiny enhancement I note there? (ideally, the message should indicate that SUL autocreating an account failed, rather than the generic account creation blocked message) [14:01:31] A bug doesn't harm. [14:02:10] The problem is that the user tries to login and receives an error about account creation which is not what was tried. [14:02:14] Right? [14:03:06] exactly [14:03:26] https://bugzilla.wikimedia.org/show_bug.cgi?id=35170 [14:03:28] there we have it [14:15:45] In SignupAPI extension, should I fire the Ajax user validation (username available & not illegal) 'onchange'/'onkeyup'? 'onchange' would cause less Ajax calls but will give result only when the user has pressed tab or clicks to the next field, while 'onkeyup' would cause a lot of Ajax calls but would give result as & when the user types so that he can correct the input if needed rather [14:15:46] than going to the next field & then finding that he just gave some invalid input on the previous field [15:30:03] akshayagarwal: regarding your earlier question - What do other sites do? [15:39:42] Reedy: most of the other sites do 'onkeyup' [15:43:19] seems that I cannot access today's or any of March's IRC log on http://ur1.ca/1e8l0 given in the channel status, should I look at some other place? [16:01:05] Parsoid just survived the expansion of [[:en:Barack Obama]] for the first time ;) [16:13:18] gwicke: well done :-]]]]]] [16:13:41] gwicke: I don't think I ever managed to have it parsed on my computers using the PHP version [16:52:26] hashar: it is not quite correct yet though.. [16:52:36] but at least it does not run out of memory any more [17:55:40] gwicke: how's it going? [17:55:54] TrevorParscal: quite well [17:56:05] that's good to hear [17:56:06] template expansion takes forever.. [17:56:26] forever to write the code, or to run the code? [17:56:34] both ;) [17:56:46] Barack Obama expanded for the first time today [17:57:00] not quite correctly, but at least it did not run out of memory [17:57:01] that sounds like a new headline [17:57:20] i recall you were shooting for something that was capable of parallel processing - is that still the case? [17:57:29] yep [17:57:38] the architecture is all prepared for it [17:57:42] nice [17:57:50] but right now the node implementation is single-threaded [17:57:55] sure [17:58:04] and not quite async either [17:58:05] are you liking working in node? [17:58:25] porting the parser to node? or everything? [17:58:34] well, it is fine for prototyping [17:58:49] and one of the better eventloop systems [17:58:54] FooBarMartijn: writing a new parser, using node.js as a prototyping environment [17:59:06] but still a bit primitive compared to GHC [17:59:15] ghc? [17:59:24] the Haskell runtime [17:59:35] ah - yes [17:59:57] but on balance I think it is a good choice [18:00:20] the language is known well enough etc [18:00:31] there's been some grumbling about it not being in PHP [18:00:45] yeah- I heard that [18:00:50] what are your thoughts on that? [18:00:50] Terry in particular I guess [18:01:00] sure, everyone has their biases [18:01:29] with a bit of googling you might find a post of me reverting a token-based parser rewrite in 2004 because it was too slow [18:01:50] and back then everything was much simpler [18:02:17] so in short- I am quite certain that the performance of a PHP implementation would suck badly [18:03:08] how much of the total CPU load is the parser btw? [18:03:22] yeah, and the trouble is that using a compiled PHP extension breaks most 3rd party use [18:03:53] dude, there are more joined/left messages than actual conversation messages in this channel [18:03:58] what is going on?! [18:04:10] people come and go so quickly here [18:04:14] srsly [18:04:47] gwicke: how stable is the HTML DOM you are emitting? [18:04:56] and is it yet documented anywhere? [18:05:34] the basic DOM is just plain HTML, but templates and so on are still in flux [18:05:46] ok [18:05:58] similarly round-trip data [18:06:11] but you should be able to just take a random HTML page for testing [18:06:17] are you imposing rules like

inside of

or are we going to have to do stuff like that ourselves? [18:06:54] I do some wrapping like that, but not inside

[18:07:12] well, the visual editor's model doesn't support

test

123

but it does support

test

123

[18:07:36] so we would have to change that, or just fake it in the conversion from HTML to linear model [18:07:56] I think so- that would also make it possible to feed it arbitrary HTML [18:08:09] not just super-sanitized DOM from the parser [18:08:24] well, writing a visual editor around arbitrary HTML is not what we are doing [18:08:49] that would easily double the complexity of the problem [18:09:03] beware of "nice-to-have"s [18:09:49] if it is not much work, then I'd definitely vote for sanitizing on the editor side [18:09:59] according to your needs [18:10:05] makes testing easier etc [18:11:20] i think what's going to happen is something like this: detect incompatible portions of the DOM tree, perform a few easy-to-reverse fix-ups, quarantine the rest into "HTML blocks" [18:11:50] eventually, we'll have to support most of what is possible in the wiki anyway [18:12:01] and that is most of HTML [18:12:03] so, if you are expecting arbitrary HTML to get round-tripped ok, sure that will work, but it's not going to be fully editable, and may be uneditable [18:12:26] we don't need a visual editor for the front-page to succeed [18:12:58] supporting everything HTML supports will supposedly eventually happen, but realistically that is a really long tail of work that will yield very little impact [18:13:06] we shouldn't expect that to happen [18:13:39] things like the

seem to be relatively minor [18:13:53] I have a hard time believing WMF is going to invest much in the long tail, history would suggest they will not [18:14:07] gwicke: sure, and that's the sort of thing we can do on the client if needed [18:14:36] do you see areas that cannot be support in principle? [18:14:41] just warning you, if you throw arbitrary HTML at the visual editor, it's not the goal of the visual editor to support editing all of it, it's the goal to support editing some of it but round trip all of it [18:14:42] *supported [18:14:56] for instance:

ab
c

[18:16:04] hmm [18:16:07] the visual editor will not make that work, because it's rediculous [18:16:51] is the problem in the data model, or in the layout? [18:16:56] but:

abc
123

could be supported by fixing it up to

abc
123

[18:17:11] the data model mostly [18:17:21] but also the UI [18:17:46] to make a sane UI, there are some things that are simply not going to work out [18:18:16] ok [18:18:36] can you give a bit more info what the issue is with the data model? [18:18:41] there are crapy situations too, like, templates that add a star to the top right of the article by making it absolutely positioned (to the nearest parent that's relatively positioned, which is the content area of the page) [18:19:05] yeah- that would need css rewriting ;) [18:19:06] so, if this template is in the editor, what should be done to edit it? [18:19:22] data model wise, we have branch and leaf nodes [18:19:25] . [18:19:45] using this model it makes it much simpler to handle a variety of complex things we have to deal with [18:20:04] branches can't have content, they can have leaf or branch children though [18:20:33] an

is a branch node, it can contain a

or a but not just raw text [18:21:14] we'll have some tension between the HTML content model and the editor's content model [18:22:23] well, if you are creating a visual HTML editor, instead of a visual wikitext editor, that would seem to be an issue [18:22:39] but if you are creating wikitext articles which have the occasional HTML in them, it's not really an issue [18:23:16] the editor will push people in a direction, away from writing out complex and arbitrary HTML, and towards clean Wikitext markup [18:23:16] if we quarantine the HTML portions that don't follow the wiki model, yes [18:23:34] it will do it this way, the corpus will gradually get cleaner [18:23:54] if we empower people to make arbitrary HTML code visually, the opposite will be true [18:24:33] well- if newly created content is clean, and old code is still editable, the you should also get a clean-up effect [18:25:01] as long as we can round trip anything, quarantine is something we can apply on a progressive basis - as we have more support we will quarantine fewer things [18:25:08] but some conversion / sanitation will likely be needed [18:25:18] could happen on-edit maybe [18:25:31] I think it should be deliberate [18:26:14] I'm open to automated cleanup and conversion tools, like a way to take a table that uses mixed syntaxes and "normalize" it with a click of a button, but it needs to be something they know they are doing [18:26:35] yeah, with a preview or something [18:26:45] if they would be unlikely to add it to a revision message (because they didn't know it was happening) it's not deliberate enough [18:26:55] but, anyways [18:26:58] good times [18:27:25] I'm going to be working with Roan on the HTML -> Linear converter [18:27:34] we will see how many issues we run into [18:27:45] cool [18:27:53] are the data attributes you are using for things pretty stable (templates aside)? [18:27:54] you can produce some DOM using parse.js [18:28:15] the data stuff is not yet completely stable [18:28:21] ok [18:28:25] no worries, just getting a feel [18:28:29] but I think I'll just keep expanding the content of data-mw-rt [18:28:40] rt? [18:28:46] round-trip [18:28:49] ah [18:28:50] yes [18:28:59] right on [18:29:21] we are working on getting you more resources [18:29:27] at least another resource [18:29:29] it has the source range, and will also have variable whitespace etc [18:30:13] I can definitely use some support ;) [18:30:17] the parser has become the bottleneck for the june "must be able to load, edit and save an existing page" target [18:30:28] not your fault [18:30:49] you have done an amazing job at getting things working quickly - we know it's been under-resourced [18:30:52] there is a reason why so many tried and failed.. [18:31:18] the finer points of the template stuff are a lot of work [18:31:48] but without getting those right, many templates won't work.. [18:31:48] yeah, people approach Wikitext and think "ah, this is simple, I can just use BNF and be done!" [18:31:53] then they realize that won't work [18:31:54] * gwicke longs for the pre-parserfunction days.. [18:32:11] if they get beyond that, they almost always say "it works great, except with templates" [18:32:41] yeah, that toothpaste will be tough to get back into the tube eh? [18:32:54] yep [18:33:06] well [18:33:18] i will let you get to it - thanks for letting me pick your brain for a bit [18:33:44] sure ;) [18:34:15] will have to read up a bit on your data model to understand more on the reasons for leaf vs. branch nodes [18:36:20] TrevorParscal: I got an inquiry from a university in London today- they might work on an alternate/faster tokenizer [18:37:22] ooh [18:37:26] sounds fancy [18:37:56] if they emit something like our tokens in XML form, we can just read that using SAX [18:38:02] and then run the same pipeline [18:41:39] FooBarMartijn: if you change something in [[:en:Barack Obama]] and then do a preview, you are mostly waiting for parsing / template expansion [18:49:28] * gwicke wonders why grepping the debug output from the expansion of a single taxobox takes so long, but then notices that it is 2.6GB ;) [18:53:12] TrevorParscal: http://dev.wikidev.net/gabriel/tmp/main_page.html [18:54:04] the result of echo '{{:Main page}}' | node parse.js [18:54:27] nice [18:54:43] no thumbnailing yet.. [18:58:29] how dare you not have that working yet! :P [18:58:45] srsly, you have made amazing progress [19:00:14] ;) [19:01:05] we need to get image dimensions anyway, so we'll likely just render them through action=parse initially [19:14:29] !c [19:14:29] See http://svn.wikimedia.org/doc/class`e1.html [19:14:32] !gc [19:14:32] https://gerrit.wikimedia.org/r/#change,`e1 [19:15:19] !class [19:15:19] See http://svn.wikimedia.org/doc/class`e1.html [19:15:23] !change [19:15:23] I don't know anything about "change". You might try: !! !api !bugzilla !cooluris !css !csshideheader !deleteuser !exclamation !gc !mainpage !mainpagetitle !package !packages !paranoid !patch !purge !rcbot !refresh !revdelete !scap [19:15:30] @dum [19:15:31] @dump [19:15:32] A dump is now available at http://toolserver.org/~mwbot/botbrain.html [19:15:47] /join mwbot [19:38:07] error I just got on mw.org: http://dpaste.com/715518/ [19:40:02] tfinc: ping [19:43:37] brb [20:39:13] Krinkle: Around? [20:40:09] Yes [20:41:55] RoanKattouw: I was away for a sec to the supermarket. Back now [20:42:00] No worries [20:42:12] I was just wondering what you were planning to work on this week [20:42:16] tfinc: pong ping [20:42:26] Krinkle: on a call [20:42:29] feel free to mai me [20:42:31] mail* [20:42:32] tfinc: ok, no rush [20:42:47] tfinc: available for a 5-10 min chat on irc within 2 hours ? [20:43:24] * tfinc looks at calendar [20:43:44] RoanKattouw: top thing right now (it's still early in the week so this may change), right now top on the list is TestSwarm. Getting special page submission into swarm to work. Doing that at an wmflabs instance [20:43:57] OK [20:44:00] i'm free 2:30 - 3 [20:44:10] Do you have any idea how long you're gonna be in testswarm/integration-land for? [20:44:23] RoanKattouw: further is code review pre 1.20 and handling of active issues in trunk regarding front-end [20:44:27] I had these fantasies about picking up RL2 again but I don't have time for it either [20:45:17] RoanKattouw: It's somewhat long-term in that the unit testing project has about 4-5 major things to do, and this is one of them. No high priority but as those things work best when done in a little sprint, I thought I'd take this week to get on that and then strike it off and leave it be for another while. [20:45:38] I'd finish the special page feature around wednessdday [20:45:44] thursday/friday RL2 ? [20:46:10] tfinc: PM, PDT I assume ? [20:46:19] Krinkle: yup [20:46:23] Hmm, that could work [20:46:27] Let's talk on Thursday [20:46:29] I am out on Friday [20:47:38] I also heard last week about the new QA lead (chrismcmahon ), proposed to meet up this week online to make sure we're on the same page. (in that the plans from last quarter for this quarter as still on) [20:47:49] hi Krinkle [20:47:50] didn't hear back yet (or I missed it) [20:47:55] chrismcmahon: did I miss it? [20:48:10] Chris was in the office last week [20:48:27] But now that he's back in Mountain Time and now that we've switched to DST, he's 2 hours closer to you [20:48:40] Krinkle: I think we both missed it. I was at WMF in SF last week, met with a LOT of people. [20:48:51] RoanKattouw: well, one hour at least. [20:48:53] I'm not in the office [20:49:28] That's what I'm saying, Chris - Europe is 7 hours now and was 9 hours last week [20:49:43] chrismcmahon: someone said we should meet up as Hashar and I are actively working on continuous integration since mid-2012 and just wanna make sure you're up to date with our plans, and get your perspective on things. [20:49:54] since mid-2011 * [20:50:15] Krinkle: I did have a phone conversation with hashar and I have some concrete ideas about how I want to use Jenkins [20:50:41] I'm referring to the mail thread "Testing on Jenkins" that Mark started [20:51:36] although the reference to Jenkins specifically and the mention of some kind of extension that came out of nowhere kind of threw me off, not sure what that is supposed to be about [20:52:37] * Krinkle adds Dashboard widget for San Francisco clock [20:52:46] :) [20:54:17] Krinkle: I sent hashar email today about the first thing I'd like to do with Jenkins (deploy a reasonable test env to labs wikis). [20:54:52] Krinkle: which "testing on jenkins" thread were you referring to ? I can not find it in wikitech-l [20:55:04] hashar: It's an internal thread [20:55:06] CC-based [20:55:13] opened by Mark [20:55:33] * hashar opens mail client: you have 387 unread emails [20:55:38] March 8, March 9 [20:55:52] chris and I on CC, also Sumana and Jeremy [20:57:29] chris^tab fail [20:57:46] :-)))) [20:58:22] Have we got a meeting in like 2 minutes? [20:58:27] chrismcmahon: what kind of test environment are you referring to? in what way would it be tied to a labs wiki (you mean beta.wmflabs.org?) [20:58:42] Reedy: yes [20:59:05] Reedy: looks like US switched DST off yesterday [20:59:08] yeah [20:59:28] see if the sip client on my phone will play nice with a bluetooth headset [20:59:40] ah, dinner saved by the minute. Gotta get it before it burns [20:59:41] perfect to me since I am back from a roughly 1000 miles trip across the country :-) [20:59:43] hashar: We switched DST /on/ yesterday, yeah [20:59:44] be back in an hour [20:59:48] Krinkle: forwarded the email to you just now [20:59:50] thx [21:00:01] I will sleep earlier this way! [21:00:08] So the time difference is temporarily 1 hour less than usual, until Europe switches (March 24th) [21:00:13] Ah, 25th [21:00:15] RoanKattouw: thanks for confirming :-] [21:00:16] for a little while [21:00:44] So yeah you guys get to go to sleep earlier this week and next :) [21:01:48] the "get to bed at 1am", "wake up at 6:30am" is not that great :-] [21:02:31] I can relate to that ;) [21:04:40] Reedy: call open [21:17:40] guillom: around? [21:18:03] hexmode: sort of, I'm doing an interview. Is there something I can help you with? [21:18:34] guillom: I'll troll around for someone else, but if you get a chance, look over https://www.mediawiki.org/wiki/Project:WikiProject_Bug_Squad [21:19:11] cool [21:20:49] Reedy: Would you have time to rerun cleanupBug31576.php against all wikis at some point this week? [21:21:38] See also https://bugzilla.wikimedia.org/show_bug.cgi?id=31576#c41 [21:22:50] Where does the script? [21:23:00] foreachwiki foobar is easy enough and just keep an eye on it [21:23:03] extensions/WikimediaMaintenance [21:23:43] Yeah... I can do it [21:23:47] Sweet, thnks [21:31:06] Krinkle: i'm back [21:31:10] Krinkle: sup? [21:32:03] tfinc: I was just wondering wether MobileFrontend to RL is still the plan. Not that it's my business but in CR it seemed like it was still under discussion [21:32:38] https://www.mediawiki.org/wiki/Special:Code/MediaWiki/113644#c32089 [21:32:41] MaxSem: [21:33:05] Krinkle: lets switch to #wikimedia-mobile [21:33:08] k [21:44:11] RoanKattouw: looks quickish [21:51:42] * gwicke sighs at http://en.wikipedia.org/w/index.php?title=Template:Taxobox/core&action=edit [21:57:46] hello [21:59:15] hi TimStarling! [22:01:00] * AaronSchulz waves TimStarling